Abstract: As a practical application of artificial intelligence in education, AI teacher essentially follows a disembodied intelligence cognitive approach, which face inherent limitations in fostering authentic epistemic engagement to generate genuine cognitive interaction and knowledge. Addressing these constraints, Rodney Brooks proposes integrating morphological intelligence with environmental affordances into robotic systems, thereby establishing a novel implementation framework to transcend disembodied limitations. Through analyzing four prototypical human-AI co-teaching models, this study reveals the evolving role of human educators. As pedagogical orchestrators, they curate instructional ecosystems while AI systems provide precise teaching support and personalized services; as collaborators, they design and implement teaching activities jointly with AI agents; as metacognitive supervisors, they primarily supervise and guide AI-led teaching processes, ensuring teaching quality and timely intervention; as adaptive counterparts, they demonstrate cognitive interoperability with autonomous AI systems. This demonstrates that AI has evolved beyond its inherent tool attributes to become an autonomous classroom participant, generating a new hybrid intelligent teaching model based on human-machine complementarity.